General Query Method

Question

I find that my C# apps do a lot of queries with a lot of boilerplate that clutters up my code space. I also want to avoid repetition, but I'm not sure how I could write a method to do this generically.

I am accessing an Oracle database using ODP. I can't use Linq because our data warehouse people refuse to designate primary keys, and ODP support for Linq appears to be, well ... they'd rather have you use their platform.

I can't really return a List because every query returns different numbers of different types.

string gufcode = String.Empty;
double cost = 0.0;
OracleCommand GUFCommand2 = thisConnection.CreateCommand();
String GUFQuery2 = "SELECT GUF_ID, COST_RATE FROM SIMPLE_TABLE";
GUFCommand2.CommandText = GUFQuery2;
OracleDataReader GUFReader2 = GUFCommand2.ExecuteReader();
while (GUFReader2.Read())
{
    if (GUFReader2[0/**GUF_CODE**/] != DBNull.Value)
    {
        gufcode = Convert.ToString(BUFReader2[0]);
    }
    if (GUFReader2[1/**COST_RATE**/] != DBNull.Value)
    {
        cost = Convert.ToDouble(GUFReader2[1]);
    }

    effortRatioDictionary.Add(bufcode, percentageOfEffort);
}
GUFReader2.Close();

But there's really a lot more terms and a lot more queries like this. I'd say 15 or so queries -some with as many as 15 or so fields returned.

Copy/pasting this boilerplate everywhere leads to a lot of fires: for example if I don't update everything in the copy paste I'll close the wrong reader (or worse) send a different query string to the database.

I'd like to be able to do something like this:

string gufQuery = "SELECT GUF_ID, COST_RATE FROM SIMPLE_TABLE";
List<something> gufResponse = miracleProcedure(gufQuery, thisConnection);

And so most of the boilerplate goes away.

I'm looking for something simple.

Your data warehouse people refuse to designate primary keys? You could have written a resignation letter in less time than it took you to post this question... — David M, Jul 19 '12 at 13:24
Oftentimes, fact tables in an OLAP schema do not need primary keys (it depends on how the data is loaded and the exact design of the schema). This is not uncommon, nor is it "wrong". I don't recommend resigning :) — EkoostikMartin, Jul 19 '12 at 13:28
My first thought is, how will you infer the types from the results. — Jodrell, Jul 19 '12 at 13:57
I guess, either you end up writng your own ORM or you could do somthing dynamic http://msdn.microsoft.com/en-us/vstudio//ff800651.aspx but tis useful to have the interface defined at compile time. — Jodrell, Jul 19 '12 at 14:05

score 1 · Answer 1 · answered Jul 19 '12 at 14:13

I think the main reason why you are not being able to abstract away a function is because the return data is going to be different everytime.

Which means the number of read is going to be different everytime as well.

You could just return GUFReader2 but then you will lose the ability to close it inside the function which you want.

I would say return an array (or list) of objects.

Inside the procedure, just read through every row and return the list while closing the connection.

Your calling function will always know what and in which sequence the expected data will be. It will have to cast the object data too, but you are doing that inside this procedure anyways.

Along with Jodrell's comment above, I decided to rethink everything I'm doing for this reason. In the past projects I've often wanted 1 dimensional lists of the same type and these methods worked great. Maybe there isn't an easy way to do this. — micahhoover, Jul 19 '12 at 17:32

score 1 · Answer 2 · answered Jul 19 '12 at 15:51

Some hints:

Deriving from IDisposable allows for cleaner code with using statement.

IMO your magic method should look more like this :

List<T> list = doMagic("SIMPLE_TABLE", columns);

columns could be an array of small structs like this one :

struct Column
{
    string Name;
    Type   DataType;
}

You might be able to use enums if you use the same tables/columns very often.

Or

You can take some inspiration from VertexDeclaration, VertexElement, VertexElementFormat and VertexElementUsage types that are in XNA : http://msdn.microsoft.com/en-us/library/bb197344.

This proven to be very helpful when dealing with a different number of 'inputs' in a random order.

In my case I've been able to build an easy to use, XNA-like framework for OpenGL with such stuff.

Regarding the return type of your list, refer to my 2nd suggestion.

score 0 · Accepted Answer · answered Jul 19 '12 at 19:38

Linq was the right answer. I give credit to David M, above, but I can't mark it as the correct answer since he only left a comment.

I was able to do a semi-generalized method using ArrayLists:

    public static ArrayList GeneralQuery(string thisQuery, OracleConnection myConnection)
    {
        ArrayList outerAL = new ArrayList();

        OracleCommand GeneralCommand = myConnection.CreateCommand();
        GeneralCommand.CommandText = thisQuery;
        OracleDataReader GeneralReader = GeneralCommand.ExecuteReader();
        while (GeneralReader.Read())
        {
            for (int i = 0; i < GeneralReader.FieldCount; i++)
            {
                ArrayList innerAL = new ArrayList();

                if (GeneralReader[i] != DBNull.Value)
                {
                    innerAL.Add(GeneralReader[i]);
                }
                else
                {
                    innerAL.Add(0);
                }
                outerAL.Add(innerAL);
            }
        }
        GeneralReader.Close();

        return outerAL;
    }

And the code that calls this method looks like this:

                thisConnection.Open();

                List<ProjectWrapper> liProjectCOs = new List<ProjectWrapper>();

                String ProjectQuery = "SELECT SF_CLIENT_PROJECT.ID, SF_CLIENT_PROJECT.NAMEX, SF_CHANGE_ORDER.ID, SF_CHANGE_ORDER.END_DATE, ";
                ProjectQuery += "SF_CLIENT_PROJECT.CONTRACTED_START_DATE, SF_CHANGE_ORDER.STATUS, SF_CHANGE_ORDER.TYPE, SF_CLIENT_PROJECT.ESTIMATED_END_DATE, SF_CLIENT_PROJECT.CONTRACTED_END_DATE ";
                ProjectQuery += "FROM SF_CLIENT_PROJECT, SF_CHANGE_ORDER ";
                ProjectQuery += "WHERE SF_CHANGE_ORDER.TYPE = 'New' ";
                ProjectQuery += "AND SF_CLIENT_PROJECT.ID = SF_CHANGE_ORDER.PROJECT";
                ArrayList alProjects = GeneralQuery(ProjectQuery, thisConnection);
                foreach( ArrayList proj in alProjects ) {
                    ProjectWrapper pw = new ProjectWrapper();
                    pw.projectId = Convert.ToString( proj[0] );
                    pw.projectId = Convert.ToString(proj[0]);
                    pw.projectName = Convert.ToString(proj[1]);
                    pw.changeOrderId = Convert.ToString(proj[2]);
                    pw.newEndDate = Convert.ToDateTime(proj[3]);
                    pw.startDate = Convert.ToDateTime(proj[4]);
                    pw.status = Convert.ToString(proj[5]);
                    pw.type = Convert.ToString(proj[6]);
                    if ( Convert.ToString(proj[7]) != "0" ) // 0 returned by generalquery if null
                        pw.oldEndDate = Convert.ToDateTime(proj[7]);
                    else
                        pw.oldEndDate = Convert.ToDateTime(proj[8]);
                    liProjectCOs.Add(pw);

There's a lot of obvious disadvantages here (although it is a lot better than what I was trying to do earlier). It is so much worse than Linq I renegotiated with our data warehouse people. There's a new guy there, and he was a lot more helpful.

Linq reduces the lines of code from above by a factor of 2. It is a factor of 4 from the non-encapsulated way I was doing it before that.

General Query Method

3 Answers3